Improved feed-forward command governor strategies for constrained discrete-time linear systems